Mixed methods
\(ATE = \lambda^Y_{01} - \lambda^Y_{10}\)
We need to
Data on exogenous variables and a key outcome for many cases
E.g., data on inequality (\(I\)) and democracy (\(D\)) for many cases
Remember: * Say we just observe a positive Inequality-Democratization correlation * Could be because Inequality causes Democratization * Could be because of confounding
Remember * Observing \(M\) helps * Process data helps address the deep problem of confounding * Key point: we don’t need \(M\) for all cases * Can learn from \(I\) and \(D\) for lots of cases and \(M\) for a subset
Suppose we go to the field and we learn that mass mobilization DID occur in Malawi
What can we conclude?
NOTHING YET!